Comprehensively Profiling the Chromatin Architecture of Tissue Restricted Antigen Expression in Thymic Epithelial Cells Over Development

Abstract

Thymic epithelial cells (TEC) effect crucial roles in thymopoiesis including the control of negative thymocyte selection. This process depends on their capacity to express promiscuously genes encoding tissue-restricted antigens. This competence is accomplished in medullary TEC (mTEC) in part by the presence of the transcriptional facilitator AutoImmune REgulator, AIRE. AIRE-regulated gene transcription is marked by repressive chromatin modifications, including H3K27me3. When during TEC development these chromatin marks are established, however, remains unclear. Here we use a comprehensive ChIP-seq dataset of multiple chromatin modifications in different TEC subtypes to demonstrate that the chromatin landscape is established early in TEC differentiation. Much of the chromatin architecture found in mature mTEC was found to be present already over earlier stages of mTEC lineage differentiation as well as in non-TEC tissues. This was reflected by the fact that a machine learning approach accurately classified genes as AIRE-induced or AIRE-independent both in immature and mature mTEC. Moreover, analysis of TEC specific enhancer elements identified candidate transcription factors likely to be important in mTEC development and function. Our findings indicate that the mature mTEC chromatin landscape is laid down early in mTEC differentiation, and that AIRE is not required for large-scale re-patterning of chromatin in mTEC.

Keywords: AIRE; chromatin immunoprecipitation; histone modifications; thymic epithelial cells; tissue restricted antigen.

Figure 1

Histone ChIP-seq samples segregate primarily by chromatin mark. (A) Correlation heatmap of histone ChIP-seq samples. (B) Principal component analysis plot of histone ChIP-seq samples. The legend shows the color both for TEC subtype (for A) and chromatin mark (for A,B).

Figure 2

Chromatin landscape around the TSS of AIRE-induced and AIRE-independent genes in mTEC^hi. Median ChIP/input signal scaled for library size is shown for each category of AIRE responsiveness (red = AIRE dependent; green = AIRE enhanced; blue = AIRE independent TRAs; purple = all other genes).

Figure 3

Chromatin landscape around the TSS of AIRE-induced and AIRE-independent genes in mTEC^lo. Median ChIP/input signal scaled for library size is shown for each category of AIRE responsiveness (red = AIRE dependent; green = AIRE enhanced; blue = AIRE independent TRAs; purple = all other genes).

Figure 4

Principal component analysis (PCA) of chromatin signatures in mTEC^lo and mTEC^hi. PCA plot of maximum signal within 1 kb of the TSS of individual genes [red = AIRE dependent; green = AIRE enhanced; blue = AIRE independent TRAs; gray = all other genes; (A) mTEClo and (C) mTEChi]. PCA rotations of individual chromatin marks for (B) mTEClo and (D) mTEChi. PCA heatmaps of genes shaded by the (E) proportion of single mTEC^hi expressing ≥ one molecule and (F) tau tissue specificity index.

Figure 5

Comparison of models to classify genes by AIRE status. Olden weightings of mTEC^hi chromatin accessibility and histone modifications within 100 neural networks for (A) all genes and (B) tissue restricted genes (tau ≥ 0.8). Accuracy (%) of neural network models generated for histone marks in common between mTEC^lo (red) and mTEC^hi (green) for (C) all genes and (D) tissue restricted genes (tau ≥ 0.8).

Figure 6

Heatmaps of RNA-seq data from TEC subtypes for significantly enriched transcription factor motifs within enhancer elements. Transcription factor motifs differentially expressed between the TEC subtypes shown are indicated by the red bars. Transcription factors expressed at FPKM > 10 and with a fold change > 5 are indicated by text to the left of the heatmap. Expression values are log₂ transformed and scaled by gene.